One hundred seventy-seven undergraduate students participated in the study (mean age: 21.1, SD: 1.65, 106 women). As an important exclusion criterion, we did not select participants with self-reported symptoms of depression (mean BDI score: 9.11, SD: 6.8). Each participant completed one experimental session and received $7 USD. Informed consent forms were signed, and the guidelines of the Code of Ethics of the World Medical Association—Declaration of Helsinki were completely followed. The ethics committees of the Universidad del Desarrollo, Pontificia Universidad Católica de Chile, and the Universidad de La Frontera approved the study at all stages.

To determine the POE categories, the DEQ was administered. From the DEQ, two indexes were obtained: dependency and self-criticism scores. Through the interaction of these factors, four categories can be generated (normal, introjective, anaclitic, and mixed A-I). Additionally, each of the participants answered the BDI to confirm the presence of self-reported depression symptomatology. In addition, the BDI's anhedonia and melancholia subscales were used for analysis. These subscales were built according to Pizzagalli et al. (17) with the anhedonia subscale considering the sum of items associated with anhedonic symptoms—loss of pleasure (item 4), loss of interest (item 12), loss of energy (item 15), and loss of interest in sex (item 21)—and the melancholia subscale considering the sum of items associated with melancholic symptoms—loss of pleasure (item 4), guilty feelings (item 5), agitation (item 11), loss of interest (item 12), early morning awakening (item 16b), and loss of interest in sex (item 21). All questionnaires were randomly administrated 1 day before the experimental protocol.

Regarding the experimental manipulation (Figure 1), participants performed a discrimination task (17, 18). After a fixation point (500 ms), a mask-face without a mouth was presented on the center of the screen (500 ms). Immediately afterward, a second face with a long (13 mm) or short (11.5 mm) mouth was presented (100 ms). Participants were asked to decide, in a separate response window, the length of the mouth (short/long), typing as fast as possible on a QWERTY keyboard (short: Q & long: W). In some trials, participants obtain a monetary reward after the decision. Participants were instructed that only a few correct answers would receive feedback. The task considers 300 trials divided into three equal blocks. Importantly, the task implemented a differential reinforcement program that objectively evaluates the tendency to change a response based on previous rewards. Thus, participants must choose between two options, only one being related to different probability of reward. One of the options is disproportionately reinforced (the rich stimulus, reinforced 30/40 of the trials per block) whereas the other is not (the lean stimulus, reinforced 10/40 of the trials per block). For half of the participants, correct responses on trials with the short mouth were the rich condition, and, for the other half of the participants, this was true for the long mouth. Given the probabilistic design, the participants could not infer which response was more advantageous based on the result of a single trial and, hence, had to consider the previous reinforcement to optimize their selections. The reinforcement procedure was carried out under a pseudo-random schedule in which it was determined which specific trial would be reinforced. If the participant did not correctly identify the stimulus in a trial intended to be reinforced, the feedback was delayed to the correct identification immediately afterwards.

Behavioral performance was analyzed using the signal-detection theory by calculating both response biases (Log-B) toward the more frequently rewarded stimulus (18, 19). Additionally, we investigate discriminability (Log-D) and performance (Response Times and Error Rate). Discriminability (Log-D) and Response bias (Log-B) were calculated considering the Pizzagalli et al. (18) formulae. (1)Discriminability:logd=12log(Richcorrect*LeancorrectRichincorrect*Leanincorrect) (2)Response bias:logb=12log(Richcorrect*LeanincorrectRichincorrect*Leancorrect) In these two formulae, Rich_correct denotes the trials in which the participant responded correctly to the condition that had an enhanced reinforcement. On the other hand, Lean_correct considered the trials in which the participant responded correctly to the diminished reinforced condition. Rich_incorrect and Lean_incorrect refer to the trials in which the participant performed erroneously for each of the rewarded condition.

First, to determine the POE categories, the DEQ was analyzed by extracting Dependency and Self-criticism factors. The categorization limits used to determine the POE groups follow the standard literature criterion (15). The samples were similarly distributed across POE groups (normal: 52, introjective: 38, anaclitic: 38 and mixed AI: 49 participants, Figure 2A). The BDI questionnaire was employed to analyze anhedonia and melancholia. First, a one-way ANOVA on melancholia scores evidenced a significant main effect of group [F_{(3, 176)} = 10.51, p < 0.01, η2 = 0.15]. Post-hoc comparisons (Bonferroni) revealed that the normal group presented significant melancholia differences with all the other groups (normal vs. anaclitic ΔM = −1.12, SE = 0.40, p < 0.05; normal vs. introjective ΔM = −1.78, SE = 0.40, p < 0.01; normal vs. mixed A-I ΔM = −1.92, SE = 0.38, p < 0.01; Figure 2C). This result on melancholia suggests that there is a differential emotional pattern between the POE configuration and the normal characterization. Regarding anhedonia, a similar ANOVA evidenced a significant group effect [F_{(3, 176)} = 5.64, p < 0.01, η2 = 0.08, Figure 2B], which was characterized by significant differences between the normal and introjective groups and between the normal and mixed AI groups (normal vs. introjective ΔM = −1.20, SE = 0.32, p < 0.01, normal vs. mixed AI ΔM = – 0.98, SE = 0.30, p < 0.01 Bonferroni). This first result confirms our hypothesis, indicating that introjective individuals, our group of interest, differ from the normal characterization in terms of anhedonia. In the same line, the anaclitic group as introjective's counterpart does not differ in anhedonia from the normal group.

Finally, to rule out the possibility that our results could be interpreted simply as depression in the participants, i.e., a confounding effect of depression on anhedonia, we repeated the analysis considering six levels of depression (16). For this purpose, we ran a two-way ANOVA on the individual anhedonia averages with fixed Depression Level and POE Group factors. The analyses show a main effect of the Depression Level [F_{(176, 3)} = 42.85 p < 0.01]; however, no interaction is observed between this factor and the POE Group (p = 0.34), which allows us to rule out the possibility that the anhedonia variability reported above is simply explained by the level of depression in general.

Results related to the experimental task show that there are no differences between POE groups' reward responsiveness and no diminished sensitivity to reinforcement in introjective individuals. Measurement of performance followed a similar pattern. First, Response Time (RT) and the mean Error Rate were analyzed using a repeated measures ANOVA with Blocks (1, 2, and 3) as within factors. POE groups (anaclitic, introjective, normal, and mixed AI) and Condition (rich and lean) were considered as between factors. We also included all possible interactions. Participants were considered the random factor. Before the analysis, mean RTs were log-transformed, and Error Rate was arcsine-transformed to approximate normal distribution. Regarding RTs, results indicated a significant main effect of Blocks [F_{(1.9, 638.4)} = 81.41, p < 0.01, η_p² = 0.19 Greenhouse-Geisser corrected] and Condition [F_{(1, 342)} = 5.03, p < 0.05, η_p² = 0.014] without any interaction among these three factors (all ps > 0.32) or with the POE main effect group (p > 0.17). Regarding performance, a significant main effect of POE group [F_{(3, 344)} = 3.38, p < 0.05, η_p² = 0.03] and an interaction between Blocks and Condition [F_{(1.62, 558.1)} = 4.33, p < 0.05, η_p² = 0.01, G-G] were observed. Post-hoc analysis (Bonferroni) showed significant differences between the anaclitic and normal groups (normal vs. anaclitic ΔM = 0.15, SE = 0.05 p < 0.05). No other effects were found (all ps > 0.37). In sum, results show that all groups improved their RTs through the task. Regarding the performance, only the anaclitic individuals showed POE group differences, improving their performance differentially through the Blocks of the task.

Second, the variables of interest were Discriminability (Log-D) and Response Bias (Log-B). The former denotes the subject's capacity to differentiate the two stimuli of the task (long or short mouth), and the latter represents the tendency to be impacted by the task reinforcement program. To test this, a repeated-measures ANOVA was run on Log-D and Log-B with factors of Blocks, POE group, and their interaction. Results for Discriminability showed a significant main effect of Blocks [F_{(1.7, 303.9)} = 13.1, p < 0.01, η_p² = 0.07, G-G corrected] with a difference between POE groups [F_{(3, 171)} = 2.69, p < 0.05, η_p² = 0.05] but without interaction between these factors (p = 0.75). In the same line, ANOVA on Response Bias again showed only a significant main effect of Block [F_{(1.8, 313.3)} = 12.7, p < 0.01, η_p² = 0.07]. No other effects were found (all ps > 0.66). Interestingly, through a linear regression, we could not show a relationship of this measure with the level of anhedonia (ps > 0.78) or with the level of depression (ps > 0.83). Consistent with the performance analysis (Error rate and RTs), Discrimination (Log-D), and Response Bias (Log-B) scores mainly increased as a function of Blocks. POE configurations were not significantly different in terms of reinforcement sensitivity (see Figure 2D).

Interestingly, the only important effect to highlight is that the anaclitic group significantly decreased its error rate in relation to the normal group. In other words, its performance increased among the blocks, and this was the POE group that performed best in the task. However, in relation to the results of Discriminability and Response Bias, this observation is not relevant in the context of our hypothesis.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.